Human Capital and Economic Growth
Robert J. Barro
Many theoretical models of economic growth, such as those of
Nelson and Phelps (1966); Lucas (1988); Becker, Murphy, and
Tamura (1990); Rebelo (1992); and Mulligan and Sala-i-Martin
(1992), have emphasized the role of human capital in the form of
educational attainment. Empirical studies of growth for a broad crosssection of countries, such as those by Romer (1990a), Barro (1991),
Kyriacou (1991), and Benhabib and Spiegel(1992), have used proxies
for human capital. These studies have, however, been hampered by
the limited educational data that were available on a consistent basis
for a large number of countries.
Recent research by Barro and Lee (1992) through the World Bank
has provided better estimates of educational attainment for a large
number of countries over the period 1960 to 1985. Hence, these data
make it possible to use a broad sample of experience across countries
and over time to assess the interplay between human capital and
economic growth. This paper summarizes preliminary empirical
results that use these data. These results provide empirical support for
economic theories that emphasize the role of human capital in the
growth process.
A new data set on educational attainment
Barro and Lee (1992) have used the census-survey data from the
United Nations and other sources for more than 100 countries. These
figures were combined with information about school-enrollment
200
Roberr J. Burro
ratios to construct a panel data set on educational attainment at
five-year intervals from 1960 to 1985. Roughly 40 percent of the cells
in this data set correspond to direct census-survey observations. The
remaining 60 percent of the cells are estimates constructed by a
perpetual-inventory method that uses the census-survey values as
benchmark stocks and the school-enrollment ratios as investment
flows.
The numbers in the data set indicate educational attainment at four
levels-no formal schooling, some elementary school, some secondary school, and some higher education-for the population aged 25
and over. This population group, rather than the labor force or the
population aged 15 and over, was dictated by the availability of data.
The figures have been used to estimate the average years of school
attainment at the primary, secondary, and higher levels. This estimation takes account of the varying duration of primary and secondary
schools across the countries and uses rough estimates of completion
percentages at these schools. It should be stressed that the estimates
do not consider variations across countries or over time in the quality
or intensity of education. The rough quality measures that are available
for a large group of countries-like measures of public spending on
education and pupil-teacher ratios-have turned out not to contribute
to the explanatory value of the human-capital variable for economic
growth or other variables.
Table 1 summarizes some major features of the data set on educational attainment. The table separates the OECD countries (22 with
data) from the developing countries, which are classed into five
regions: Middle EastNorth Africa ( I 4 countries), Sub-Saharan Africa
(27 countries), Latin AmericaICaribbean (23 countries), Pacific Area
(10 countries), and Other Asia (7 countries). The population figures
shown are for the overall population of the region, although the
schooling data apply to those 25 and over. The figures on educational
attainment show the average years of schooling at the primary, secondary, and higher levels, and the total of these three categories. The
regional averages were formed as unweighted means of the individual
country observations.
Human Capital and Economic Growth
Table 1
Trends of Educational Attainment by Region
RegionIGroup Year
OECD
(22 countries)
1960
1965
,1970
1975
1.980
1985
-Middle East/
North Africa
414)
1960
,1965
.I970
1975
,1980
1985
Sub-Saharan
Africa
(27)
1960
1965
.I 970
Total
Population
(millions)
Average years of schooling in
primary
secondary
higher
total
Latin America1 1960
Caribbean
1965
(23)
1970
1975
1980
1985
Pacific
Area
(10)
1960
1965
1970
1975
1980
1985
Other
Asia
(7)
1960
1965
1970
1975
1980
1985
Note: Attainment applies to the population aged 25 and older, but the population
figures shown are for total population. The regional values are unweighted means
of the average number of years of schooling in each country.
202
Robert J. Barro
The table shows that the OECD group had the highest school
attainment, beginning with 6.2 total years in 1960 and reaching 8.3
years in 1985. The developing regions have, however, all grown faster
in proportionate terms and have therefore been catching up in average
years of schooling to the OECD countries. The lowest attainment is
in Sub-Saharan Africa, with a range from 1960 to 1985 of 1.2 to 2.2
years, whereas the highest is in the Pacific area, with a range of 3.3 to
5.3 years.1 (Some of the countries in this group-Hong Kong, Singapore, Korea, and Taiwan-now have sufficiently high per capita
income so that they no longer warrant the designation of developing
country.)
Human capital in theories of economic growth
Various theoretical models include human capital as a factor of
production and assess the accumulation of human capital as an element
of the growth process. I consider first the role of human capital in the
familiar neoclassical growth model, then examine the implications of
theories that allow for imbalances between human and physical capital. Human capital is also important in models that allow for international capital mobility and in theories of the diffusion of technology.
Finally, I assess the interplay between human capital and choices of
fertility rates.
The convergence rate in the neoclassical growth model
The standard framework that often guides economists' thinking
about economic growth is the neoclassical growth model of a closed
economy, due to Ramsey (1928), Solow (1956), Cass (1965), and
Koopmans (1965). The long-run per capita growth rate in this model
depends entirely on the exogenous rate of technological progress. In
the short run-that is, in the transition to the steady state-the growth
rate depends inversely on the gap between economy i's per capita
product or income, denoted by yi, and its long-run or steady-state
position, denoted by yf .2 This result is often referred to as conditional
convergence: economy i grows faster the ~owkrits initial income, yi,
conditional on its long-run target, yf. In the standard mode1,yf
depends positively on the economy's willingness to save and level of
Human Capital and Economrc Growth
203
productivity and negatively on the population growth rate. In extended
versions of the model, the effective level of productivity can be
interpreted to include not only the access to technology, but also
government policies in regard to taxation, maintenance of property
rights, provision of infrastructure services, and so on.
The transitional dynamics can be summarized by the rate of convergence: how much of the gap between yi and yf is eliminated in one
year? Empirical evidence discussed by Barro and Sala-i-Martin (199 1,
1992a) for the U.S. states (from 1880 to 1988), regions of seven
Western European countries (from 1950 to 1985), and a cross-section
of about 100 countries (from 1960 to 1985) indicates that the rate of
convergence is on the order of 2 percent per year. That is, if the
differences across economies in yf are held constant, then about 2
percent of the gap between the typical poor and rich economy is
eliminated in one year. This slow rate of convergence means that it
takes 35 and 1 15 years, respectively, for 50 percent and 90 percent of
the initial gap to vanish.
For the regions of the United States and Western Europe, the
steady-state values, yf, appear to be similar, and hence, conditional
convergence corresponds to the poor economies catching up to the
rich ones. For the broad group of countries, however, the variations in
the yTappear to be substantial, partly because of persisting differences
in government policies. In this context, therefore, conditional convergence does not imply that the poor countries would tend to grow faster
per capita than the rich countries.
In the neoclassical growth model, the convergence rate depends
mainly on the speed with which diminishing returns to capital set in.
If yi is well below yT-so that the ratio of capital to labor, ki, is well
below its steady-state value, kf-then the rate of return on capital is
high and the economy tends to grow rapidly. As the economy
develops, yi and ki rise, the rate of return on capital falls, and the growth
rate tends to decrease.
If capital is viewed narrowly-say to include machines and buildings but to exclude human capital-then the share of capital in income
204
Robert J. Burro
would be low, diminishing returns to capital would set in quickly, and
the convergence rate would be high.3 It therefore turns out to be
infeasible (if we assume plausible values for the various parameters
in the model) to reconcile the neoclassical growth model with a narrow
concept of capital. The model fits much better with the empirical
estimates of convergence speeds if we take the appropriately broad
view of capital to include huinan components. A capital share of about
three-quarters-a reasonable figure if human capital is includedgives a slow enough onset of diminishing returns so that the theory
can generate a convergence rate of about 2 percent per year. Thus, the
slow observed rates of convergence provide indirect evidence for the
importance of human capital accumulation in the process of development.
Imbalances between physical and human capital
Extensions of the neoclassical growth model have distinguished the
sector that produces goods--consumables and physical capital-from
an education sector that produces new human capital (see, for example,
Lucas [I9881 and Mulligan and Sala-i-Martin [1992]). The assumption
in these models is that the education sector is relatively intensive in
human capital: it takes human capital embodied in teachers to produce
human capital in students.
One finding stressed by Mulligan and Sala-i-Martin (1992) concerns
imbalances between human and physical capital, that is, departures of
the ratio of human to physical capital from the ratio that prevails in
the long run. The key result is that a higher ratio of human to physical
capital and hence, a higher ratio of human capital to output raises the
growth rate. A country with an abundance of human capital tends also
to focus its investment on physical capital; that is, a high ratio of human
to physical capital results in a high ratio of physical investment to gross
domestic product.
The conclusions about imbalances between human and physical
capital are reinforced if the accumulation of human capital involves
adjustment costs that are much higher than those applicable to physical
capital. (Machines and buildings can be assembled quickly, but people
cannot be educated rapidly without encountering a sharp falloff in the
Human Capital and Economic Growth
205
rate of return to investment.) An economy with a high ratio of human
to physical capital is then like an economy that is described by the
transitional dynamics of the usual neoclassical growth model. The
economy effectively starts with a quantity of physical capital per
worker that is substantially below its steady-state position, that is, far
below the amount that matches the large quantity of human capital.
The usual convergence effect implies that the growth rate of output
exceeds its steady-state value in this situation.
A high ratio of human to physical capital applies, as an example,
after a war that destroys large amounts of physical capital, but which
leaves human capital relatively intact. Japan and Germany after World
War I1 are illustrative cases. The theory accords with the empirical
observation that countries in this situation tend to recover r a ~ i d l y . ~
Capital mobility
The discussion thus far assumes a closed economy: goods do not
move across borders, and the residents or government of one economy
cannot borrow from or lend to those in another economy. This assumption is unrealistic for countries, but is especially troubling for the
analysis of regions of the United States or the Western European
countries.
It is possible to extend the neoclassical growth model to allow for
international trade in goods and assets (see, for example, Barro and
Sala-i-Martin [1992b, Ch.21). One result from this extension is that
the opening up of the economy to world credit markets speeds up the
predicted rate of convergence to the steady state. This speeding up
applies especially to forms of physical capital that are not subject to
adjustment costs and that can be financed by international borrowing
(or are amenable to direct foreign investment). If all capital were of
this form and if international credit markets were perfect, then a small
country's capital stock and production would converge essentially
instantaneously to the steady state.
Human capital provides little collateral for lenders and therefore
typically cannot be financed by borrowing (or direct foreign investment). Hence, even in an open economy, the accumulation of human
206
Robert J . Barro
capital must be financed primarily with domestic savings. This linkage
between domestic investment and domestic saving restores the key
assumption of the standard neoclassical growth model for a closed
economy: capital is subject to diminishing returns and at least part of
the capital stock must be financed by domestic savings. The bottom
line turns out accordingly to be that the open-economy model with
human capital generates rates of convergence that are only slightly
higher than those of the standard neoclassical model. If the share of
broad capital-physical plus human-is around three-quarters, then
the predicted rates of convergence can still match the observed values
of about 2 percent per year.
The di8usion of technology
The most interesting aspect of the recent literature on endogenous
economic growth, represented by Romer (1990b) and Grossman and
Helpman (1991, Chs. 3,4),concern theories of technological progress
in the leading economies. In these models, a technological advance
shows up either as the discovery of a new type of product (a new kind
of productive input or a new variety of final good) or as an improvement in the quality or productivity of an existing product. These
advances require purposive research effort, although the output from
the research sector may involve random elements.
The incentive to commit resources to research requires a reward for
success. In the models, the rewards take the form of monopoly rentals
on product innovation. That is, a successful innovator's monopoly
position lasts for awhile because of first-mover advantages, secrecy,
and possibly formal patent protection.5
Growth can be sustained in these models if diminishing returns do
not apply, that is, if the returns from new discoveries do not decline
in relation to the costs of making the discoveries. One reason that
diminishing returns may not apply is that the potential supply of new
ideas and products is effectively unlimited.
For a single economy, the endogenous technological progress
generated in recent theoretical models substitutes for the exogenous
technological progress that is assumed in the standard neoclassical
Human Capital and Economic Growth
207
growth model. For studying convergence across economies, the interesting application of the new theories is to the pirocess of adaptation
or imitation by followers of the innovations that were made by leaders.
The cost of imitation for a follower can be modeled as similar to the
cost of discovery for a leader, except that the cost of imitation is likely
to be smaller and subject to less uncertainty. These considerations
suggest that a follower would grow faster than a leader and thereby
tend to catch up to the leader. This conclusion may not hold, however,
if the follower country's environment is hostile to investment (in the
form here of expenses for technological adaptation) because of poorly
defined property rights, high rates of taxation, and so on.
Although innovation in the world economy may not be subject to
diminishing returns, the process of imitation by a single country would
encounter diminishing returns as it exhausts the pool of innovations
from abroad that are readily adaptable to the domestic context. This
consideration leads to the usual convergence property: a follower
country tends to grow faster the larger the stock of potential imitations
and hence, the further its per capita income is from that of the leaders.
The convergence result is again conditional on aspects of the domestic
economy-such as government policies, attitudes about saving, and
intrinsic levels of productivity-that affect the returns from technological adaptation.
Nelson and Phelps (1966) pointed out that a country with more
human capital would be more adept at the adaptation of technologies
that were discovered elsewhere. Thus, the higher the stock of human
capital for a follower country, the higher the rate of absorption of the
leading technology and hence, the higher the follower country's
growth rate.6 This conclusion resembles the one that we got from
imbalances between the stocks of human and physical capital; each
model predicts a positive relationship between the initial stock of
human capital per person and the subsequent per capita growth rate.
Human capital and fertility
In the standard neoclassical growth model, a higher rate of population growth reduces the steady-state value of capital per worker and
Roberf J . Barro
thereby lowers the steady-state value of per capita income, Y!. The
decrease in yF implies that the economy grows in the transition (for a
given value of yi) at a slower rate. The rate of population growth is
exogenous in this model, and the effect on the steady-state level of
capital per worker involves the flow of new capital that has to be
provided to accompany the flow of new workers.
Richer theories, such as the one by Becker, Murphy, and Tamura
(1 990), include the resources expended on children and allow fertility
to be a choice variable of families. A key result is that a larger stock
of human capital per person raises the wage rate and therefore the time
cost of raising children. (The assumption is that the productivity in the
sector that raises children does not rise as fast as that in the sectors that
produce goods and new human capital.) A higher stock of human
capital motivates families to choose a lower fertility rate and to raise
the investment in human capital for each child (that is, to substitute
quality for quantity in children). These responses of population growth
and human capital investment tend to raise the growth rate of output.
This model therefore provides another channel through which a larger
stock of human capital results in a higher subsequent rate of economic
growth.
Empirical evidence on human capital and growth
across countries
Table 2 contains a sample of empirical results from ongoing research
on the effects of a number of variables on the growth rate of real per
capita GDP. (The data on GDP are the purchasing-power-parity
adjusted values constructed by Summers and Heston [1988].) The
estimates apply to a panel data set for 73 countries-those with a full
set of data-over five-year periods from 1960 to 1985. There are 365
observations in total, five time observations for 73 countries. The
estimation is by the seemingly unrelated (SUR) technique, which
allows the error term for each country to be correlated over time.
The independent variables include the logarithm of real per capita
GDP at the start of each period, l~g(yit),a number of variables
including government policies that can be interpreted as determinants
209
Human Capital and Economic Growth
of a country's steady-state position, y:, and the educational-attainment
variable. See the notes to Table 2 for details.
Table 2
Panel Regressions for Growth Rate of Real Per Capita GDP,
5-Year Intervals from 1960 to 1985
Independent
Variable
log (Initial GDP)
Estimated Coefficients & Standard Errors
-.0167
(.0027)
-.0196
(.0024)
-.201
(.101)
-.050
(.085)
-.0226
(.0054)
-.0208
(.0049)
-.0147
(.0074)
-.0107
(.0062)
log (School)
Openness*log (l+Tarriff
Rate)
log (l+Black-Market
Premium)
Freq. of Revols. and Coups
FERT
Sub-Saharan Africa
--
--
Latin America
--
--
.05, .38,
.22, .31,
.08
.07, .52,
.26, .44,
.22
R ~indiv.
,
periods
Roberr J. Barro
Notes to Table 2
The dependent variable is the annual growth rate of real per capita GDP
over each period (1960-65, 1965-70, 1970-75, 1975-80, 1980-85). These
data are from Summers and Heston (1988). Standard errors are shown in
parentheses. There are 365 observations (73 countries and 5 time periods).
Coefficients are estimated by seemingly unrelated (SUR) technique, which
allows a country's error term to be correlated over time. Separate constants
are estimated for each time period. Other coefficients are constrained to be
the same for all periods.
Initial GDP is real per capita GDP at the start of each 5-year interval.
School is 1 plus the average number of years of educational attainment
for the population aged 25 and over at the start of each 5-year period.
G/Y is the period average of the ratio of real government consumption,
exclusive of education and defense, to real GDP.
Openness is an estimate of "natural" openness, based on area and
distance measures. This variable is a constant for each country.
Tarlff rate is an average of official tariff rates on capital imports and
intermediates, weighted by shares in imports. Only one observation per
country was available for the tariff rate.
Black-market premium is the period average of the black-market
premium on foreign exchange.
Frequency of revolutions and coups is the number of revolutions and
coups per year, averaged over the full sample, 1960-85.
I/Y is the ratio of real gross domestic investment to real GDP, averaged
over each period.
FERT is the total fertility rate, averaged over each period.
Sub-Saharan Africa is a dummy for countries in sub-Saharan Africa.
Latin America is a dummy for countries in Latin America.
For given values of the other variables, the estimated coefficient on
log(yit), in the first regression is -.0167, s.e. = .0027. Thus, this
coefficient differs significantly from zero (t-value = 6.2), and the
magnitude indicates a rate of convergence to the steady-state position
of 1.7 percent per year.7
The determinants of yTcontained in the first regression of Table 2
are G/Y, the ratio of government consumption exclusive of education
and defense to GDP, a measure of distortions due to tariff^,^ the
black-market premium on foreign exchange-intended as a proxy for
2//
Humart Caprrul and Economic Growth
distortions in foreign trade,9 and the frequency of revolutions and
coups-intended as a proxy for political stability. These variables
affect growth in the expected manner in the first regression: all have
negative effects on the growth rate. Since these variables are not the
major concern of the present paper, I will not provide a detailed
assessment of these results.
The schooling variable is entered as log(l+total years of school
attainment), where the years of attainment apply to the start of each
period. The parameter 1 in the above expression can be viewed as the
'~
effective number of years obtained without formal s c h ~ o l i n g .The
estimated coefficient on the schooling variable in the first regression,
.0232, s.e. = .0041, is positive and highly significant (t-value = 5.7).
Thus, for a given value of log(yi,), and for given values of the
determinants of yT, countries grew faster if they began each period
with a greater amount of educational attainment. As a quantitative
example, if average educational attainment begins at two years-the
average value prevailing in Sub-Saharan Africa in 1980-theri an
increase by 0.3 years would raise the quantity, 1 + years of attainment,
by 10 percent and thereby increase the predicted growth rate by 0.2
percentage points per year. (The effect diminishes gradually over time
because logbit) then follows a higher path than it would have otherwise.)
The second regression shown in Table 2 adds In,the ratio of real
gross domestic investment to real GDP, and the total fertility rate.
(These variables are measured as averages over each period.) In the
Solow version of the neoclassical growth model, the investment ratio
(or the saving rate) and the fertility rate (or the growth rate of
population) are exogenous variables. These variables do not influence
the long-run growth rate, but do affect the steady-state level of per
capita output, yT. An increase in Inraises yT, whereas a rise in fertility
lowers yT. Therefore, for a given value of logbit), an increase in I/Y
would raise the growth rate, whereas an increase in the fertility rate
would lower the growth rate.
~ r o man econometric standpoint, the exogeneity of I/Y and the
fertility rate with respect to the growth rate are questionable. In any
''
212
Robert J. Burro
event, the second regression in Table 2 shows that the estimated
coefficient of H i s positive and highly significant (.120, s.e. = .021),
whereas that for fertility is negative and significant (-.0037, s.e. =
.0012). These results are consistent with the Solow model of economic
growth.
For present purposes, the most interesting finding from the second
regression is that the inclusion of the investment ratio and the fertility
rate roughly halves the estimated coefficient on the schooling variable:
the estimated value is now .0109, s.e. = .0041. This result suggests that
a good deal of the effect of initial human capital on the growth rate
works through its effects on investment and fertility. These channels
of effect are examined below.
The third and fourth regressions shown in Table 2 include dummy
variables for Sub-Saharan Africa and Latin America. Both continent
dummies are significantly negative, substantially so for Sub-Saharan
Africa. The main inference from these results is that the variables
considered thus far-including the estimate of educational attainment-are insufficient to explain a significant part of the poor growth
performances in these regions. One possibility is that the measures of
educational attainment in Sub-Saharan Africa, although low (see
Table l ) , do not fully capture the low levels of human capital in this
region.
Table 3 shows regressions in the same form as Table 2 for the
investment ratio, I . , and the total fertility rate. These variables are
measured as averages over the periods considered. For present purposes, the important findings are that the schooling variable has a
significantly positive effect on I/Y in the first two regressions and a
significantly negative on the fertility rate in the last two regressions.
Thus, these results confirm the idea that part of the influence of initial
human capital on the growth rate involves the positive interaction with
investment in physical capital and the negative interaction with the
fertility rate. The interaction with physical investment would occur,
for example, in the model of imbalances between human and physical
capital that was worked out by Mulligan and Sala-i-Martin (1992).
The interplay with fertility arises in the theory of Becker, Murphy, and
Tamura (1990).
Human Capital and Economic Growth
213
The results shown in the second regression of Table 2 showed that
the effect of the school-attainment variable on the growth rate
remained significantly positive even after holding constant the investment ratio and the fertility rate. A possible interpretation, along the
lines of Nelson and Phelps (1966), is that this effect of human capital
reflects the enhanced ability to adapt new technologies.
Concluding observations
Economic theory suggests that human capital would be an important
determinant of growth, and empirical evidence for a broad group of
countries confirms this linkage. Countries that start with a higher level
of educational attainment grow faster for a given level of initial per
capita GDP and for given values of policy-related variables. The
channels of effect involve the positive effect of human capital on
physical investment, the negative effect of human capital on fertility,
and an additional positive effect on growth for given values of investment and fertility.
Ongoing research is considering the possibilities for improving the
measures of educational attainment, especially by using better data on
enrollment ratios and more information about school dropouts. The
possibilities for measuring the quality of school input, in addition to
the quantity, are also being considered.
School attainment is, in any event, only one aspect of human capital.
Another dimension is health status. Measures of life expectancy-a
proxy for health status-turn out to have substantial explanatory value
for economic growth and fertility; life expectancy at birth enters in a
way similar to educational attainment in the regressions reported in
Tables 2 and 3. The interplay between health capital and educational
capital is currently being investigated.
Robert J. Burro
214
Table 3
Panel Regessions for Ratio of Real Investment to Real
GDP and Total Fertility Rate, 5-Year Intervals
from 1960 to 1985
Indevendent Variable
IN
Fertility Rate
Estimated Coefficients and Standard Errors
log (initial GDP)
log (School)
Openness*log
(1+Tariff Rate)
log (1+Black-Market
Premium)
Freq. of Revols. and
Coups
Sub-Saharan Africa
Latin America
R ~indiv.
, periods.
Note: The dependent var~ablefor the first two regressions is the average over each period of
the ratio of real gross domestic investment to real GDP (data from Summers and Heston
[1988]).For the last two regressions, the dependent variable IS the average over each period
of the U.N. estimate of the total fertility rate (average number of live births per woman over
her lifetime). See also the notes to Table 2.
Human Capital and Economic Growrh
Endnotes
'Table 1 does not cover the formerly centrally-planned economies. These countnes had
average years of schooling that were similar to the OECD countries.
%he quantities y, and y: have to be Interpreted as values filtered for the effects of exogenous
technological progress. The usual procedure is to compute output per unit of effective labor.
where effective labor is the aggregate amount of work effort multiplied by the cumulative effect
from labor-augmenting technological change.
3 ~ hconvergence
e
rate depends also on whether the saving rate falls or rises as an economy
develops. If a poor economy saves a lot and then lowers its savlng as it grows, then the
convergence rate would be hlgher, and vlce versa. Solow (1956) assumed aconstant saving rate,
and the optimizing models(of Cass [I9651and Koopmans [1965]) that allow for a varylng saving
rate make no clear predictions about whether the rate will fall or rise as an economy develops.
(The falling rate of return suggests that the saving rate would decline, but the rise in Income
toward its permanent level suggests the opposite.)
4 ~ Imbalance
n
In the other dlrectionMahlgh ratio of physical to human capital, perhaps as a
consequence of an epidemlcUcanalso lead to a growth rate that exceeds the steady-state growth
rate. The effect of this klnd of imbalance on the growth rate would be relatively weak, however,
if the accumulat~onof human capital were subject to large adjustment costs.
5 ~ h ipaper
s
focuses on the role of these models as positive theories of economlc growth and
abstracts from the Inferences that have been drawn for desirable governmental policies. The
policy implications derive from positive or negative gaps between social and private rates of
return. Pos~tivegaps can reflect uncompensated spillover benefits In research and production,
the consequences of monopoly pricing of the existing goods, and the disincentive effects from
taxation. Negativegaps can come from the seeklng of exlstlng monopoly rentals by new entrants
or from congestion effects (negative spillovers from economic activity).
?he stock of human capital would also tend to reduce the cost of innovation in leading
economies. Hence, more human capital can speed up the rate of innovat~on,an effect that raises
the growth rate in leading and following economies.
'More precisely, because the estimation IS canied out at five-year intervals, the coefficient,
,0167, has to be adjusted slightly to compute the instantaneous rate of convergence (see Barro
and Sala-i-Martin [1992a]). The implied convergence coefficient turns out in this case to be 1.8
percent per year.
he tariff rate enters as an interaction wlth an estimate of natural openness, the country's
Thls
ratio of imports to GDP that would have occurred in the absence of trade d~stort~ons.
openness was estimated to be a negative funct~onof the country's area and 11sweighted-average
distance from major markets The Idea IS that distortions due to tariffs have a larger adverse
influence on growth for countries that are naturally more open (small countries and countries
that are close to major potentlal trading partners). See Lee (1992) for a discussion
he black-market premium may also proxy more broadly for otherdistortionary policies and
for macroeconomic instablllty.
I 0 ~ h value
e
1.0 IS close to the non-hear, maximum-likel~hoodestimate of this parameter In
Robert J. Bnrro
the form of the first regression shown In Table 2. The value was then restncted to 1.0 and was
not reestimated for the vanous regressions shown. The logarithm~cform used in the regresslons
turned out to fit sllghrly better than a linear form In attainment.
he emplncal results are slmllar, however, if lagged values of //Y and FERT are used as
instruments. The exogeneity of other vanables In the regresslons. such as revolutions and coups
and the black-market premlum, can also be questioned.
References
Barro, R. J. "Economic Growth in a Cross-Secclon of Countnes," Quarrerl~ Jourrtn/ of
Economics. 106 (199 I), pp. 407-43
and J. Lee "Internat~onalCompansons of Educational Attainment, 1960-1985,"
unpubllshed manuscnpt, Harvard Un~versity.June 1992.
and X Sala-i-Manln. "Convergence Across States and Regions." Brookrttgs Pnpers
on Economic Acrrvlr?.,No. 1 (1991j, pp. 107-58.
and
. "Convergence," Jortrttal of Polirrcal Ecor~orn~,
100 ( 1 992a) pp
223-5 1.
and
. "Economic Growth," unpubl~shedmanuscript, Harvard University, 1992b.
Becker, G. S , K. M. Murphy. and R. Tamura. "Human Capltal. Fenlllty, and Economlc Growth,"
Jourrtal of Polrrrcal Ecottot?t?; 98 (October 1990 supplement), pp S 12-37
Benhabib, J . and M. M. Spe~gel."The Role of Human Capital and Poht~calInstability in
Economlc Development." unpubllshed rnanuscnpt. New York Unlverslty, May 1992.
Cass, D. "Optlmum Growth In an Aggregate ,Model of Capltal Accumulation," Ret.ie&t'of
Econotnic Studies, 32 (July 1965). pp. 233-40.
Grossman. G M. and E Helpman. I r ~ t t o ~ ~ ~artd
r t oGron.th
t~
rtt theGlobrrl E c o t t o n ~Cambndge,
~.
Massachusetts M.I.T. P r e ~ s1991.
,
Koopmans, T.C. "On the Concept of Optlmal Economic Growth." In The Econot?tetrrcAppronch
ro De~elopmenrPlannrrtg. Amsterdam: North Holland. 1965
Kynacou. G. A. "Level and Growth Effects of Human Capital. A Cross-Country Study of the
Convergence Hypothesis," unpubllshed manuscnpt, New York University, June 1991.
Lee, J. "International Trade, Distortions. and Long-Run Economlc Growth," unpublished
manuscript, Harvard Unlcerslty. 1992
Lucas, R E. "On the Mechamcs of Development Planning." Jortrturl ofMottetan Econot?tics,
22. (July 1988). pp. 3-42.
Mulllgan, C. B and X. Sala-I-Martin. "Some Emplncal Evldence on Two-Sector Models of
Endogenous Growth," unpubllshed manuscnpt, University of Chicago, June 1992.
Nelson, R. R. and E. S . Phelps. "Investment in Humans, Technological Diffusion, and Economic
Growth," American Ecottot?lic Re~,rewPapers crt~dProceed~rtgs,56. (May 1966), pp. 69-75.
Ramsey, F.P. "A Mathematical Theory of Savlng." Ecottot71rcJourt~al,38, (December 1928) pp.
69-75.
Rebelo, S. "Long-Run Pollcy Analysis and Long-Run Growth." Jourr7ol of Polrricol Econont~
99 (June 1991). pp 500-21.
Romer, P. M "Human Capital and Growth. Theory and Evidence," Cnrrtegie-Rochester Cottferettce Series on Prrbltc Policy 1990a.
Solow, R M. "A Contribution to the Theory of Economic Growth." Quanerl? Jo~rnlnlof
Ecottomrcs, 70 (February 1956). pp. 65-94.
Summers. R. and A. Heston. "ANew Set of Internat~onalComparisons of Real Product and Price
Levels Estimates for 130 Countnes. 1950-1985." Review of Irtcome and Wealth 34 (March
1988). pp 1-25.